Debugging WebRTC Calls

Reporting WebRTC Call Issues

The best way to report an issue is through Bugzilla using this link. Describe the issue you’ve run into, include a URL, along with the details of the call setup. See the Adding Call Setup Information section for helpful templates. Here are some common examples of descriptive WebRTC bug summaries:

  • caller is experiencing frozen video, screen capture, or desktop capture

  • caller does not hear audio

  • caller’s voice sounds distorted or robotic

  • video resolution is lower than expected

  • caller’s video appears rotated

  • there is significant delay between a caller’s video and audio

  • camera, microphone, or screens are not appearing in the Firefox device access permission prompts, etc.

  • caller’s video is garbled, partially missing, or the colors are incorrect

  • caller can not share external display, but can share integrated display


Not all web conferencing software makes extensive use of WebRTC.

For simple issues, the first place to look is to check the web developer console for error messages related to media format issues. If you see messages here related to WebRTC, getUserMedia, or getDisplayMedia, please add this information to your bug.

Adding Call Setup Information

The following template can help one provide all the call details needed to diagnose many common issues.

* Does this problem occur in Firefox for Desktop or Android?

* Is this problem reproducible in Firefox Nightly?

* Has this worked previously?

* Have you tried using `about:profiles` to reproduce the problem in a
  clean profile?

* How many participants were in the call?

* In which web conferencing or web calling services does the problem occur?

* Does the problem present itself immediately upon starting the call?

* If not how long does it take to occur?

* If this is a problem with audio or video capture, what camera or microphone
  are you using? (adding about:support text may be helpful)

* If this is problem with screen capture, which screen was being captured,
  and are there other screens attached to the same machine?

* Would you be willing to help us find a regression range?

If the issue is of specification compliance the template provided below may be more useful. If one is unsure if this is a compliance issue, one can refer to the Standards Documentation section for links.

* What unexpected behavior is being observed? What are the steps to reproduce
  this issue?

* What is the expected behavior?

* Where is this behavior specified?

* Is this problem reproducible in Nightly?

* Have you tried using `about:profiles` to reproduce the problem in a clean

* Has this worked previously?

* If so, would you be willing to help us find a regression range?

Adding about:support Text

In your Bugzilla report, please include support information about the current device on which you are experiencing an issue.

  1. Open a tab and visit about:support

  2. Click ‘Copy Text to Clipboard’

  3. Paste this text in your Bugzilla bug comment and post.


To open about:* links in Firefox one must do the following: #. Right-click the link and select Open Link in New Tab #. Select the new tab #. Click inside the address bar which should contain the about URL #. Press Enter

Adding about:webrtc RTCPeerConnection Stats

  1. Open about:webrtc.

  2. Expand the RTCPeerConnection section.

  3. Locate and expand the subsection RTCPeerConnection that one wishes to copy stats from.

  4. Press the Copy Stats Report

  5. In the Bugzilla bug, press the Attach New File button.

  6. Click inside the large text box labeled File, and paste the copied stats report.

  7. Add a descriptive label to the the Description:, e.g. “PeerConnection statistics sample taken during period of frame loss”.

  8. In the drop down box, next to the select from list radio option, select “JSON source (application/json)”.

  9. If needed, add a descriptive comment to the comment field.

  10. Press the Submit button.


Hovering the mouse over some headings will reveal a clipboard icon. Clicking this icon will copy the contents under that section to the clipboard as a JSON document. This can be useful if one wants to submit a portion of the available stats. Of particular note are the RTP Stats heading whose button will copy the latests RTP stats, and the SDP section whose button will copy the SDP offer; answer; and roles.

Adding Your about:webrtc Contents

For issues with call quality, please share web conferencing related performance information by providing your about:webrtc information. Note this information should be collected while the call in question is still active.

  1. While your call is still ongoing, open an additional tab and visit about:webrtc.

  2. Click “Clear History” to clear the stats from other recent calls which are no longer ongoing.

  3. At the bottom of the page click ‘Save Page’, and save this file.

  4. Add this file as an attachment to your bug.

This data contains statistics about your call, the signalling that was used to setup your call, and information about the network transports.

Diagnosing Call Quality Issues

about:webrtc Overview

about:webrtc is an in-browser resource for debugging WebRTC calls. The primary audience for about:webrtc is browser developers but it can also be of use to anyone that needs to troubleshoot a WebRTC call. When there is no call data to display, about:webrtc will appear as follows:

about:webrtc before any calls have been made

about:webrtc before any calls have been made

Note that there are several sections. During a call each section will contain information pertinent to different aspects of the WebRTC browser implementation.

RTCPeerConnection Statistics

This section presents information useful for diagnosing active calls. It contains RTCPeerConnection creation parameters, connection information, negotiation details, RTP stream statistics, bandwidth statistics, and output frame statistics.

Connection Log

When one needs to diagnose issues with the underlying transports, logs can be found under Connection Log.

User Modified WebRTC Configuration

This section will display any user modified preferences which effect the performance or behavior of browser components which can impact WebRTC calls. When hovering over a preference path displayed in this section a clipboard icon will appear. Clicking that icon will copy the path to the clipboard. It can then be pasted into about:config, to change or reset the value to its default.


Unexpected values in this section may be caused by installed extensions. It is best to test issues in a clean profile when possible using about:profiles.

Media Context

Information that is gathered to determine Codec availability and capability is recorded under Media Context.

Bottom Control Bar

At the bottom of about:webrtc is a row of buttons which allow the user to perform miscellaneous actions.

The Save Page button expands all the sections and presents a dialog to save the contents of the page. This will produce an HTML file suitable for attaching to bug reports.

In the event of a WebRTC issue, the Enable WebRTC Log Preset button is a very quick way to start logging. Pressing that button will open a new tab with about:logging with the webrtc preset selected. This preset contains all Standard Logging Modules. Logging will begin immediately. If one needs to change other log settings on that page one can customize them, and then press Start Logging. This may be necessary if one wishes to log to the profiler.

If experiencing echo cancellation issues, one may be asked to submit echo cancellation logs. These logs are gathered by pressing the Start AEC Logging button. One should press the button to activate the logging while actively experiencing an echo cancellation failure.


Producing echo cancellation logs is incompatible with the content sandbox. The user will be prompted with further instructions if the sandbox is active and the Start AEC Logging button is pressed.

Audio/Video Delay

Delayed media is commonly caused by long physical paths between endpoints, though anything that slows down inter-hop delivery of packets can be at fault. Note that this is different than the bandwidth of the network path, and a high latency will not be fixed by reducing the video resolution or audio sample rate. Round trip time, or RTT, is the time it takes for a packet to get from the sender to the receiver and then for a packet to get from the receiver back to the sender. If the path is symmetric between the two endpoints one can assume that the one way delay is half the delay of the round trip.

The second common cause of A/V delay is jitter, the magnitude of variability in packet inter-arrival times. In order to smoothly play out of the incoming stream a receiver experiencing jitter will have to buffer (delay) incoming packets.

Using about:webrtc to Diagnose Delay

The key metrics in about:webrtc are RTT (round-trip-time) and jitter. They can be found in the RTP stats section of the PeerConnection. The PeerConnection informational blocks start out in a minimized state, and one will need to expand a block to find the RTP stats section:

How to reveal the full statistics of a PeerConnection in about:webrtc

How to reveal the full statistics of a PeerConnection in about:webrtc

Once expanded one can locate the RTP Stats section and find the key diagnostic stats:

Location in about:webrtc of jitter and RTT stats

Location in about:webrtc of jitter and RTT stats

In this image we can see that there are 0 milliseconds of jitter, and 32 milliseconds of round trip delay. This call should not be experiencing any noticeable delay. See the Delay Calculation appendix section below for some more detail.

If the perceived delay is larger than the estimated delay that could indicate a problem within Firefox that requires debugging. Under these circumstances it would be helpful to grab a JSON copy of the current stats by pressing the “Copy Report” button, pasting those stats into your Bugzilla bug report.

Location in about:webrtc of Copy Report button

Location in about:webrtc of Copy Report button

Performance Profiling and Logging

Capturing a Firefox Performance Profile

For basic performance issues, a performance profile can help engineers diagnose issues with video formats, performance, and rendering.

  1. Visit and enable the Profiler toolbar button.

  2. Click the toolbar button down arrow and select ‘Media’ in the Settings drop down.

  3. Open a tab and visit the page with the affected media content.

  4. Click the Profiler toolbar main button to start recording.

  5. Play media until the issue you are seeing manifests.

  6. Click the Profiler toolbar button again to stop recording.

  7. When a new Profile tab opens, click the upload profile button on the upper right.

  8. Copy the resulting profile URL and post this to your Bugzilla report.

Additionally, detailed logging can be collected within performance profiles to help aid in debugging complicated issues. To enable the collection of a profile with low level debugging -

  1. Visit and enable the Profiler toolbar button.

  2. In a new tab, visit about:webrtc. Click the ‘Enable WebRTC Log Preset’ button, which will open a tab for about:logging with pre-populated information.

  3. In about:logging, click the “Start Logging” button. (You are now recording a profile, the profiler toolbar toggle button should be selected automatically.)

  4. Open a new tab for testing and view the media you are having an issue with. (After reproducing, DO NOT close this test tab yet.)

  5. Switch to the about:logging tab, click ‘Stop logging’, and then close the test tab.

  6. Wait approximately 10 - 20 seconds for a new tab to automatically open containing the generated performance profile.

  7. Within the upper-right side of the profiler tab click the ‘upload local profile’ button to initiate profile upload. Once the upload is complete, a drop down text field will open displaying the URL of the profile. Select this text and copy it.

  8. Share the URL of the profile for analysis with the engineer you are working with.

Alternatively one can set the following environment variable:


Note that webrtc_trace will not be active until “Enable WebRTC Log Preset” is pressed.

Standard Logging Modules

Standard Logging Modules







JSEP state machine



SDP parsing



Network transports



JS API related to receiving media and media control packets



JS API related to sending media and media control packets



JS API related to sending DTMF messages





media capture

Content process end of IPC channel for receiving frames from media capture devices


media capture

Parent process end of IPC channel for sending frames from media capture devices


media capture

Orchestrates capture of frames from media capture devices in the parent process


media capture

Object pool of shared memory frame buffers for transferring media capture frames from parent to child process


media capture

Captures tab content for sharing































Glue code between transport, media, and libwebrtc components



implements the RTCPeerConnection object




libwebrtc logging

Prior to Firefox v123 it must be enabled from about:webrtc at runtime, or it must be set in the MOZ_LOG environment variable at launch.



implements the RTCRtpTransceiver object




Non-standard Logging Modules

Standard Logging Modules







Logs RTP and RTCP packet fragments

See Debugging Encrypted Packets

Examining Call Performance Issues

Enabling Call Stats History

Call stats history is enabled by default in Nightly. To enable in release builds open about:config, and change “media.aboutwebrtc.hist.enabled” to true. This will keep a history window of stats for a number of recent calls, allowing for inspection in about:webrtc after a call has completed.

Dumping Call Stats

One can dump a JSON blob of call stats for an active call, or a recent call if call stats history is enabled. There are two buttons in about:webrtc to do this, “Copy Report” and “Copy Report History”. The former will create a copy of the most recent stats for the PeerConnection. The later will copy all the history of stats reports that about:webrtc has accumulated for that PeerConnection, this can be up to several minutes of stats.

Debugging Encrypted Packets


There is a blog post covering dumping unencrypted partial RTP and RTCP packets in the logs. While the information presented in that post is still relevant, the command to extract the packet data in the blog is out of date. A working method is presented below;

Using the gecko logging system, unencrypted, mangled, partial, RTP-packets can be written out. This may be a good avenue of investigation for packet loss and recovery, simulcast, and feedback. Because the entirety of the packet is not guaranteed to be logged, this is less suitable for debugging issues with encoded media. These logged packets can be converted to PCAP files, which can then be explored in Wireshark. The logs produced by this module can be quite large, making it easy to identify by file size which child process log files contain packet dumps.

To start RTP logging, one must enable the RtpLogger log module. The sync option should also be used, as it prevents undesirable interleaving of log messages. Here are the minimal log settings needed:


In order to interpret the packet contents, one needs to refer to the SDP. Wireshark is unaware of the negotiated details, so it can not directly decode the media, nor can it decode the header extensions. The SDP can also be logged, and so the following is a more useful set of log settings:



On macOS it is simple to install Wireshark and text2pcap with Homebrew:

# Use only one of the following:
# ==============================

# To install the Wireshark GUI application and the command line utilities:
brew install --cask wireshark

# To install only the command line utilities:
brew install wireshark

One can use tee to capture log output from a copy of Firefox launched from the command line, e.g. through mach. Alternatively, one can set a log file through the environment variable MOZ_LOG_FILE or through about:logging.


If log files are not being created by child processes, this is likely due to sandboxing of content processes. To work around this one must either select a location within the sandbox, disable the content sandbox, or launch Firefox from the command line, e.g. from a Firefox dev environment:

MOZ_LOG=sync,RtpLogger:5,jsep:5 MOZ_LOG_FILE= ./mach run 2>&1 | tee your.log

To produce a PCAP file one needs to filter the logs to include only the RtpLogger log lines, reduce them down to the expected ingestion format for text2pcap, and finally to invoke text2pcap.

cat your.log  | rg 'RtpLogger.*RTC?P_PACKET|>>\s(?P<packet>.+$)' --only-matching  --replace '$packet' | text2pcap -D -n -l 1 -i 17 -u 1234,1235 -t '%H:%M:%S.' - your.output.pcap


If rg, a.k.a ripgrep, is not already available, one can install it via one of the following methods:

# Install through cargo on macOS, Linux, or Windows
cargo install ripgrep

# Install via Homebrew on macOS
brew install ripgrep

# ripgrep packages may be available through the package manager for your
# Linux distro

The resulting PCAP file can be explored with Wireshark. Currently, one must refer to the SDP in order to interpret the RTP packets.

# On most Linux distros
wireshark -d 'udp.port==1234,rtp' your.output.pcap

# On macOS when installed via Homebrew
open /Applications/ --args -d 'udp.port==1234,rtp' your.output.pcap

Examining Codec Availability and Capabilities

When codec negotiation doesn’t happen as expected there are several helpful locations where one can find information. The SDP offer and answer contain the list of codecs that were in the initial offer, and the subset of those codecs that were selected in the answer.

The easiest way to get this information on a live call is through about:webrtc. Each RTCPeerConnection has its own subsection, that when expanded contains an SDP section. There are buttons to display the offer and the answer. Depending on which party was the offerer and which was the answerer one may have a local offer and a remote answer, or a remote offer and a local answer.

Firefox chooses which codecs to offer based on availability. Some codecs, like Opus or VP8, are always available. Some codecs are available in software and some codecs on some platforms are available in hardware. H264 support is provided by a third-party, and is automatically downloaded the first time its use is requested. This is a process which can take a variable amount of time depending on network circumstances.


A list of media codecs with playback support are available in the Media section of about:support#media . Not all media codecs present and available to Firefox for playback are supported in WebRTC calls.

To check the current factors, including preferences, that are being used to calculate availability beyond codec presence, one can check the Media Context section of about:webrtc.

example about:webrtc media context values

For an in-depth reference covering the codecs available through WebRTC please see the MDN Page: Codecs Used by WebRTC.

Running WebRTC Tests

There are a variety of tests providing coverage over WebRTC related code. The Web Platform Suite provides conformance tests for browsers. The gtest suite is composed of unit tests. Crashtests are a type of regression test which are written to induce crashes. There are fuzzing tests which exercise APIs in ways that the authors did not foresee. All of the WebRTC tests can be run locally with mach or in CI on Try. There is a detailed overview of all available test types including those not exercised by WebRTC code here.


Running ./mach <verb> --help is an indispensable tool for discovering options that can streamline your testing process.


A test suite on Try maybe an aggregate of multiple logical test suites. For example, the mochitest-media suite on try includes both the WebRTC and playback mochitests.


WebRTC calls make use of a number of internal timers. Amongst the behaviors these timers control are transport selection, bandwidth estimation, packet loss determination, media adaptation, lip sync, connection timeout, and more. There are Try targets which are too slow to reliably run a number of the tests. Before running a specific test on Try for the first time, it may be best to check the relevant test suite manifest. This can be done easily with by searching for and viewing a test file. If that test has been disabled on one or more platforms the details will appear as shown below: warning that the displayed test file has been disabled on Android

Test Atlas

WebRTC Test Locations


Test type

Test file location

Try suite

Treeherder Abbreviations





mda, M(mda)

Web Platform Test



wpt, W(wpt)





WebRTC Signalling





WebRTC (gUM / gDDM)

Browser Chrome Test (mochitest)



bc, M(bc)

WebRTC Transport







SDP parser




Web Platform Tests

The WPT suite comprises conformance tests for various W3C specs such as: CSS, JS APIs, and HTML. WebRTC is a JS API and as such its tests are of the testharness.js type. There is detailed WPT documentation available here Web Platform Tests can be run locally from

# Run the entire WebRTC WPT test suite
./mach wpt testing/web-platform/tests/webrtc

# Run a single test, e.g. RTCPeerConnection-createAnswer.html
./mach wpt testing/web-platform/tests/webrtc/RTCPeerConnection-createAnswer.html

# Run all of the PeerConnection tests, i.e. RTCPeerConnection-*.html
# NOTE that the `mach` verb in use is `test` not `wpt`
./mach test testing/web-platform/tests/webrtc/RTCPeerConnection


Running the WPT tests locally can be very disruptive to one’s working desktop environment, as windows will frequently appear and grab focus. Using mach’s --headless flag will prevent this, and can be a great way to run them if one’s problem can be evaluated from command line output.

These tests are synced from the main Web Platform Test repository, and likewise our changes are synced from our in-tree copy back to that repository.


Running the WebRTC mochitests in Try is done using the entire Web Platform Test suite, wpt. As such, this can be slow.

./mach try fuzzy --query 'wpt'

One can run those same tests in Chromium, Safari, or Servo if one needs to compare behavior between browsers. This can be done directly through mach, see running tests in other browsers for more details.


The WebRTC mochitests are integration tests, regression tests, and sanity tests. The needs of these tests did not align with specification conformance testing in the WPT, Web Platform Test, suite. Before writing a new mochitest, one should consider if a test would be better expressed as a WPT, which all browsers can test against.

Locally running the WebRTC mochitests should be done in a Firefox dev environment using mach as follows:

# Run the whole suite
./mach mochitest dom/media/webrtc/tests/mochitests

# Run a single test, e.g. test_peerConnection_basicAudioVideo.html
./mach mochitest dom/media/webrtc/tests/mochitests/test_peerConnection_basicAudioVideo.html
# Or
./mach mochitest test_peerConnection_basicAudioVideo.html

# Run all of the PeerConnection tests, i.e. test_peerConnection_*.html
./mach mochitest test_peerConnection

On try, WebRTC mochitests are part of the larger media test suite. This suite can be easily selected with the following fuzzy query:

# Run the media mochitest suite on all regular platforms
./mach try fuzzy --query 'mochitest-media'

# Run the media mochitest suite only on Linux which will resolve far faster
./mach try fuzzy --query 'linux mochitest-media'


The gtests are all compiled into a single library target: xul-test. This makes running gtests from mach slightly different than the other test types.

# Run a single test by using Prefix.TestName, e.g. JsepSessionTest.FullCall
./mach gtest 'JsepSessionTest.FullCall'

# Run all the tests in a single Prefix, e.g. JsepSessionTest
./mach gtest 'JsepSessiontTest.*'

# Run all tests which have a Prefix.TestName containing the substring 'Jsep'
# See the table of selectors below
./mach gtest '*Jsep*'

# Run all the gtests for Firefox
./mach gtest

Here is a list of helpful substring selectors for executing specific WebRTC gtests:

WebRTC GTest Selectors





JSEP (signalling) tests

jsep_session_unittest.cpp jsep_trak_unittest.cpp


SDP parsing tests



MediaPipline and MediaPipeline filter tests for RTP media handling



AudioConduit tests for libwebrtc glue for RTP audio media



VideoConduit tests for libwebrtc glue for RTP video media


For more general information about gtests see the documentation here.

Fuzz Testing

It is not common to need to run fuzz testing as it is run on a semi-continuous fashion in CI. It is more likely that one will need to respond to a bug filed by a fuzzing bot. If one is interested in fuzzing one should consult the excellent documentation available here.

Code Atlas

There are a number of components that work together to create a successful WebRTC call. When debugging a call it can be difficult to see the larger puzzle for all the pieces. A listing of the WebRTC related source code directories is provided below to help one navigate.

WebRTC Code Atlas







This is the primary directory for Firefox WebRTC code



This contains WebRTC related utility code



This contains the C++ implementations of the JavaScript WebRTC interfaces



This is the JSEP state engine implementation


WebRTC (various)

This is the glue code between libwebrtc and Firefox



This contains the SDP parsing interface



This contains some of the WebRTC related tests



The scripting and configuration for vendoring new versions of libwebrtc are here

This is unlikely to be of concern for debugging



This contains the ICE implementation, the MDNS implementation, and transport code



This contains the MediaPipeline and MediaPipeline filter code which is glue between transport and the libwebrtc RTP stack



This is the SRTP implementation used by Firefox


WebRTC (various)

libwebrtc handles many aspects of WebRTC calls above the transport layer and below the presentation layer



webrtc-sdp is a Rust implementation of a WebRTC-only SDP parser



sipcc is a C implementation of a general SDP parser

this carries many local modifications


Media Capture

GetUserMedia and related classes are here

There are many other unrelated media source files here



This contains the WebIDL definitions for the WebRTC JS API amongst many other WebIDL definitions


Standards Documentation

When debugging API behavior it may be necessary to consult the specifications for WebRTC. The ECMAScript API is defined in several W3C standards, webrtc-pc, and webrtc-stats. The number of IETF standards that are incorporated into WebRTC are too numerous to list here. One can find these standards in the Normative References section of the webrtc-pc spec.

Appendix: Delay Calculation

For all intents and purposes jitter and RTT are additive in nature. If there was 25ms of jitter reported and a RTT of 272ms, one could estimate the expected delay from transmission at the send side to play out on receive side to be

25ms + (272ms / 2) = 161ms